M ETHODOLOGICAL CONSIDERATIONS

The strengths and limitations of the methods used in this project and the different papers are discussed in the following section.

5.2.1 Design & measurements

Detailed protocols were generated prior to the start of the study to limit extraneous variables and enhance control over test procedures and data collection. Three of the studies (Papers II, III & IV) had a cross-sectional design. This design is appropriate for describing relationships among phenomena when there is a cogent theoretical rationale behind the analysis [137]. We investigated the relationship between neck pain and aspects of dizziness and balance, which is founded on theoretical theories and previous evidence between interactions of the cervical afferent system, the visual and the vestibular system. The advantage of such a design is that it provides better precision and control of the data collection, enhancing the precision of the association in question. A drawback of this broad methodological approach is the inability to gain a deeper understanding of the results with the largest disadvantage being the lack of ability to conclude in a causal way. This makes it difficult to consider the internal validity of the project as it cannot establish causal effects [138].

Using subjective measures such as questionnaires as outcome measures could introduce bias as the patient may not be in the necessary physical or psychological state to give accurate opinions of their experienced health status or be influenced by recall bias. They may also be concerned by the consequences of their answers in terms of care given by the healthcare provider[139, 140]. However, patient-reported outcome measures is a valuable tool as it provides insight into the patient’s

perception of their own health, which is important information to providing patientcentered care [140]. In addition, the questionnaires used in this thesis are widely used and have previously been validated. As this was a multicenter study, there were more than one assessor examining the patients and conducting the physical

tests. Several assessors on the different centers may have led to measurement bias.

Thus, prior to the study, the examiners had two sessions and then an additional session after five months for calibration of the different tests to ensure consistency in the measures. The physical tests used in Paper IV have previously shown adequate validity and reliability and were assessed by experienced physiotherapists. In addition to the use of validated questionnaires, this improves the quality of the data and that it measures the intended construct, which is important when considering the internal validity of the project [139].

In Paper II, the diagnostic process was thorough and carried out by an otolaryngologist; however, a large portion of patients were diagnosed with a nonvestibular diagnosis. Even though the diagnostic process was based on several objective measures, the study could have been strengthened by the inclusion of an objective measure of vestibular function, such as the caloric test. Paper III used posturography as a measure of postural sway. Even though it is a widely used tool and indicated to be reliable, the findings by Ruhe et al [89] indicated that at least three trials should be used with 90 seconds of data acquisition. This deviated from our protocol and we acknowledge that the results should be interpreted with caution.

Further, we used PPT as a way of measuring pain sensitivity in the neck, which has been proposed to affect cervical afferent input [48, 141]. The PPT has shown good reliability and concurrent validity when compared to other subjective measures of pain [80]. For these reasons, PPT seems beneficial and feasible for research purposes.

However, PPT does not directly measure altered proprioception of the neck. Perhaps other tests that are directly aimed to measure proprioception should have been added, such as the joint position error [142].

5.2.2 Setting and sample

There are limitations to consider when interpreting the results and considering the external validity and thus, the generalizability of this thesis. The population may be prone to sampling errors as the inclusion criteria were fairly wide and based on referral

for either dizziness or neck pain to a specialized care unit. Thus, the populations were heterogeneous as we included patients based on symptom complexes and not a specific diagnoses. Caution should thus be exercised when considering the generalizability of the studies. Stricter inclusion criteria for neck pain could have improved the generalizability of the results, making the sample less heterogeneous. For instant, using subgroups of patients with neck pain as proposed by Guzman et al. [143]. However, subgrouping would lead to a reduction in group sizes and thus statistical power.

The inclusion of patients with these symptom complexes in this setting is also a strength. The associations were examined in two unselected patient groups with dizziness and neck pain, i.e. the patients were not selected due to any a priori assumption of a causal link between their neck symptoms and dizziness or balance.

The results from this thesis may be generalizable to these types of patients, referred for either dizziness or neck pain, as they appear in a clinical setting in a specialized care unit. Women were overrepresented at both centers; however, women are usually overrepresented in both neck pain and dizziness populations [2, 144].

In Paper III and IV, the populations recruited from the outpatient spine center did not undergo an otoneurologic examination of their dizziness. Even though the diagnosis was not the objective of these studies, an overview of the patients having vestibular dysfunction in this group would perhaps enhance the clinical value and the interpretation of the studies.

One strength of this thesis was the relatively large sample size. However, the sample size was a convenient sample and a power estimation was only conducted for Paper III.

There was a relatively large difference in group sizes in Papers III and IV, which should be considered when interoperating the results from these studies. The results from Paper IV call for caution as several associations were investigated, increasing the risk of type I errors.

5.2.3 Systematic review

The results and comparison of the different studies in the systematic review (Paper I) should be considered with caution as the test procedures, equipment and variables differed across the studies. A considerable limitation was the low number of studies, varying outcomes and the relatively low methodological quality of the included studies, making pooling of data and meta-analysis not possible. However, the lack of clinical studies on patients with CD emphasizes the importance of increasing the knowledge in this field. In addition to the small numbers of studies, a limitation of the systematic review was that half of the studies were more than 9 years, old with publications dating from 1993 to 2017. Thus, the results from this review should be interpreted with caution. In addition, this review only reflects the diagnostic criteria for the studies meeting its inclusion criteria and is thus not representative of all studies on CD. However, the fact that the included studies had to have a comparison group for clinical outcome would probably not exclude other valuable clinical studies on patients with CD. Another limitation could be the inclusion of studies comparing patients with CD with healthy controls which makes the review somewhat

heterogenic. However, as we do not know how these patients differ from other diagnoses or even healthy controls, and with the general low number of clinical studies examining these patients, we found that comparisons with healthy controls would contribute to the limited knowledge within the area. The strength of this review is the thorough and systematic search process, adherence to guidelines and the use of two independent reviewers.